A field guide to picking your first automation

The first automation a business builds sets the tone for every one after it. Get it right and the team starts spotting candidates on its own. Get it wrong – too ambitious, too vague, too clever – and "automation" becomes the word for that project that broke and got switched off. So choose carefully, and choose small.

The three tests

Pick something that happens every day, takes real time, and has a clear right answer. Every day, because a weekly task never builds enough momentum to feel worth it. Real time, because ten minutes a day is an hour a week and that is a number people notice. And a clear right answer, because this is the one that decides whether the project works at all: if two people can disagree about whether the task was done correctly, it is not a first automation. It might be a third one.

Measure it before you touch it

Hours per week, errors per month, how long the queue gets on a bad day. Write the numbers down and date them. A baseline is what turns an automation from a nice idea into a decision you can defend, and it is the difference between "it feels faster" and "it saves eleven hours a week" when someone asks whether it was worth doing. Almost nobody does this. Everybody wishes they had.

Ship the smallest version

The version that removes the work, and nothing more. No dashboard, no notifications, no configuration screen. Run it for two weeks with the old process still available. Watch what it gets wrong, fix that, and only then widen it. The temptation to build the complete system at once is strong, and it is how first automations become last ones.

Good first candidates

Copying orders from one system into another. Renaming and filing the documents that arrive by email. Chasing the same status update from the same supplier every week. Turning a form submission into a ticket, a row and a reply. Checking that yesterday's stock file actually arrived, and shouting if it didn't. Every one of these is dull, and every one of them is a day back a month.

Bad first candidates

Anything with the word "intelligent" in the pitch. Anything customer-facing, where a mistake is seen by someone who can leave. Anything that replaces a judgement call rather than a repetitive one. And anything the business does not do yet – automating a process that has never run manually means automating a guess.

The first automation is not there to change the business. It is there to prove the business can change, and to teach everyone what good looks like. Once that is in place, the second one is easy to find. It is usually the thing the first one revealed.